Analysis Of Protein Spots Detection Approach In Two Dimensional Gel Electrophoresis Images

Two-dimension gel electrophoresis (2-DE) is one of the important researchtechnologies in proteomics，and the valuable information extraction from differentialproteins is the core of2-DE. Facing the complex images of2-DE, how to detect proteinspots automatically, rapidly and effectively is the difficulty and the emphasis of theinformation extraction. This paper studies the preprocessing and protein spot detectiontechnologies in2-DE images, and presents several related methods and algorithms forpractical applications. The main research results are concluded as follows:(1) The preprocessing method of2-DE images is studied. Aiming at thephenomenon that2-DE images are easy to subject noise disturbances caused by theprocess of electrophoresis, dyeing and scanning, this paper applies fast median filtering,Gaussian filtering and morphological filtering methods for image de-noising. For theinhomogeneous background of2-DE images, this paper uses morphological top-hattransform method to estimate the background value. Experiments on real2-DE imagesshow that these preprocessing methods obviously reduce the impact of noise andbackground in protein analysis, and improve the accuracy of detection.(2) The Laplacian of Gaussian-based algorithm for spot detection in2-DE imagesis studied. First, compute the Laplacian of the smoothed image, and store the Laplaciandirection and magnitude values in two additional images. Second, extract all of spotregions by neighborhood connectivity properties and a variety of heuristic methods indetermining spot extents. Finally, normalize the valid spot quantifications. Experimentson real2-DE images show that this algorithm is very sensitive to noise, although it candetect most protein spots in PDQuest standard gel images and good quality cameraimage, but poor to gel-scanned images with noise serious, inhomogeneous backgroundand serious over-saturated.(3) The marker-controlled watershed transformation-based algorithm for spotdetection in2-DE images is studied. In order to solve the over-segmentation andwatershed lines discontinuous problems of the traditional watershed algorithm，thispaper presents an improved second watershed transformation-based algorithm, inconjunction with morphological grayscale reconstruction and distance transformation to extract and control regional minimum. In order to extract valid spots, this paper definesa qualified inspection standard to reduce the impact of false minimum. Experiments onreal2-DE images show that this algorithm can detect most protein spots in PDQueststandard gel images, camera image and gel-scanned images, and can resolve thewatershed lines discontinuous and over-segmentation problems. The qualified inspectionstandard is an effectual means for valid spots extraction.